Antipsychotic medication and suicidality

Our results demonstrate that, when compared to periods of non-use, use of antipsychotics was associated with 70 and 69% lower risk of attempted or completed suicide, and readmission with suicidal plans, respectively. This is in accordance with previous research, where non-adherence to antipsychotic drugs consistently emerges as a significant risk factor for suicidality in individuals with schizophrenia (Hassan et al., Reference Hassan, De Luca, Dai, Asmundo, Di Nunno, Monda and Villano2021; Hor & Taylor, Reference Hor and Taylor2010; Pompili et al., Reference Pompili, Amador, Girardi, Harkavy-Friedman, Harrow, Kaplan and Tatarelli2007; Taipale et al., Reference Taipale, Lähteenvuo, Tanskanen, Mittendorfer-Rutz and Tiihonen2021; Tiihonen et al., Reference Tiihonen, Wahlbeck, Lonnqvist, Klaukka, Ioannidis, Volavka and Haukka2006b). Active hallucinations and delusions are established risk factors for suicide in individuals with schizophrenia (Hor & Taylor, Reference Hor and Taylor2010). Moreover, insomnia has been proposed as a contributing factor to suicidal ideation, suicide attempts, and completed suicide within this group (Waite, Sheaves, Isham, Reeve, & Freeman, Reference Waite, Sheaves, Isham, Reeve and Freeman2020). Antipsychotic medication has demonstrated efficacy in reducing hallucinations and delusions in patients with schizophrenia (Sommer et al., Reference Sommer, Slotema, Daskalakis, Derks, Blom and van der Gaag2012), and antipsychotics can also enhance sleep (Monti, Torterolo, & Pandi Perumal, Reference Monti, Torterolo and Pandi Perumal2017). These properties may contribute to the protective effect of antipsychotics on suicidality.

Antidepressants and suicidality

We found no associations between the use of antidepressants and the risk of neither attempted or completed suicide, nor readmission with suicidal plans. While prior studies on combination therapy in patients with schizophrenia have indicated an impact of antidepressants on negative symptoms (Singh, Singh, Kar, & Chan, Reference Singh, Singh, Kar and Chan2010) and all-cause mortality (Haukka et al., Reference Haukka, Tiihonen, Harkanen and Lonnqvist2008; Lin, Yeh, & Pan, Reference Lin, Yeh and Pan2023), few studies have demonstrated a significant effect on suicidality (Reutfors et al., Reference Reutfors, Bahmanyar, Jönsson, Brandt, Bodén, Ekbom and Osby2013; Tiihonen et al., Reference Tiihonen, Lonnqvist, Wahlbeck, Klaukka, Tanskanen and Haukka2006a, Reference Tiihonen, Suokas, Suvisaari, Haukka and Korhonen2012). The evidence regarding the association between antidepressants and suicide risk is somewhat conflicting (Ballon & Stroup, Reference Ballon and Stroup2013; Dragioti et al., Reference Dragioti, Solmi, Favaro, Fusar-Poli, Dazzan, Thompson and Evangelou2019). Although some previous studies have suggested that antidepressants are associated with an increased short-term risk of suicidal acts in young people with psychiatric disorders (Hawkins et al., Reference Hawkins, Coryell, Leung, Parikh, Weston, Nestadt and El-Mallakh2021), the majority of studies, including comprehensive systematic reviews (Dragioti et al., Reference Dragioti, Solmi, Favaro, Fusar-Poli, Dazzan, Thompson and Evangelou2019), have concluded that antidepressants represent a safe treatment option for psychiatric disorders.

Benzodiazepines and suicidality

Our results revealed a significant association between the use of prescribed benzodiazepines and increased risk of readmission with suicidal plans, but not with increased risk of attempted or completed suicide in patients with schizophrenia. However, previous studies have identified associations with risk of attempted and completed suicide in this patient group (Dodds, Reference Dodds2017; Tiihonen et al., Reference Tiihonen, Suokas, Suvisaari, Haukka and Korhonen2012, Reference Tiihonen, Mittendorfer-Rutz, Torniainen, Alexanderson and Tanskanen2016). One potential explanation for the discrepancy between our results and previous studies may lie in the infrequency (n = 32) of attempted and completed suicides within our study cohort. We cannot exclude the possibility that this is a type 2 error, due to lack of statistical power. Moreover, confounding by indication is a well-known challenge in pharmacological cohort studies, as the use of medications is not randomized. Aggression and agitation not only represent risk factors for suicide (McGirr & Turecki, Reference McGirr and Turecki2008; Pompili et al., Reference Pompili, Lester, Grispini, Innamorati, Calandro, Iliceto and Girardi2009), but also serve as common rationales for the prescription of benzodiazepines in patients with schizophrenia (Dold et al., Reference Dold, Li, Tardy, Khorsand, Gillies and Leucht2012). Many prior studies on suicidality and benzodiazepines have failed to adjust for the patients' levels of aggression and agitation, and it is not surprising that these studies identify an association between benzodiazepine use and suicide risk. In our study, we addressed this concern by adjusting for the baseline levels of overactive, aggressive, disrupted, or agitated behavior. This adjustment might, at least partly, explain why our findings differ from those reported in previous studies.

Other findings

In line with other studies on suicidality in patients with schizophrenia (Hor & Taylor, Reference Hor and Taylor2010; Sher & Kahn, Reference Sher and Kahn2019), we found that younger age was independently associated with increased risk of attempted or completed suicide. Interestingly, our results did not reveal a significant association between age and the risk of readmission with suicidal plans, suggesting that while suicidal ideation spans across all age categories, the occurrence of actual suicides and suicide attempts is most prevalent among younger patients within a psychiatrically hospitalized cohort such as this. With regard to gender, a notable difference in suicidal behavior has been documented in previous literature, revealing a higher prevalence of females with non-fatal suicidal incidents and a predominance of males in completed suicides, commonly referred to as the ‘gender paradox of suicidal behaviour’, just as in the general population (Schrijvers, Bollen, & Sabbe, Reference Schrijvers, Bollen and Sabbe2012; Sher & Kahn, Reference Sher and Kahn2019). However, in patients with schizophrenia, previous research has indicated that the gender gap in completed suicides is notably less pronounced compared to the general population (Sher & Kahn, Reference Sher and Kahn2019). In line with this, our results showed that female gender was associated with an increased risk of being readmitted with suicidal plans. Moreover, no association was found between male gender and increased risk of attempted or completed suicide, perhaps because the primary outcome of the analyses included both attempted and completed suicides, not solely completed suicides.

Numerous studies have identified depression, agitated behavior, and abuse of alcohol and substances as potential risk factors for suicide in patients with schizophrenia (Hettige et al., Reference Hettige, Bani-Fatemi, Sakinofsky and De Luca2018; Hor & Taylor, Reference Hor and Taylor2010; McGirr & Turecki, Reference McGirr and Turecki2008; Sher & Kahn, Reference Sher and Kahn2019). In contrast, our results show no significant associations between the risk of attempted or completed suicide and either depressive mood, overactive, aggressive, disrupted, or agitated behavior, nor excessive use of alcohol or illicit substances. Similarly, no such associations were found between these factors and the risk of readmission with suicidal plans. This can possibly be attributed to the fact that our analyses relied solely on assessments of these factors at the initiation of the index admission. However, depression, agitation, and abuse of alcohol and substances are known to fluctuate over time, limiting the usefulness of one initial measurement.

Strengths and limitations

A key limitation of this study is its observational design, involving the non-randomized use of psychotropic drugs. This limits our ability to draw causal conclusions and raises the risk of confounding by indication. Confounding by indication occurs when clinical indications influence treatment decisions and are independent risk factors for the study outcome (Sendor & Stürmer, Reference Sendor and Stürmer2022). Benzodiazepines are commonly prescribed to patients with schizophrenia to manage agitation, aggression, or anxiety (Dold et al., Reference Dold, Li, Tardy, Khorsand, Gillies and Leucht2012), all of which are also risk factors for suicidality (McGirr & Turecki, Reference McGirr and Turecki2008; Pompili et al., Reference Pompili, Lester, Grispini, Innamorati, Calandro, Iliceto and Girardi2009). This confounding by indication may, at least partially, explain our observed association between benzodiazepine use and an increased risk of readmission with suicidal plans. Similarly, antipsychotic drugs are prescribed to manage psychosis (Tandon, Nasrallah, & Keshavan, Reference Tandon, Nasrallah and Keshavan2010). Since psychotic symptoms are well-known risk factors for suicidality, our finding of an association between antipsychotic use and a reduced risk of suicidality may actually be an underestimation. However, to address the issue of confounding by indication, we have accounted for well-known confounding variables in our analyses. Another possible approach to address this issue could involve excluding, for instance, the first 30 days of drug exposure. Unfortunately, conducting such an analysis was not feasible without compromising the statistical model used in this study.

Other limitations include the following: First, instances of suicidality that did not result in an acute psychiatric admission were not recorded. While the impact of incorporating such data into our analysis remains uncertain, it is likely that similar trends would persist, even among less severe manifestations of suicidality. Second, we lacked data regarding the doses of psychotropic medication used, and accordingly, we were unable to analyze the potential impact of varying doses on the risk of suicidality. Third, the term ‘non-use’ of antipsychotic drugs included both non-adherence and clinician-guided discontinuation. Given the assumption that discontinuation under a clinician may be associated with lower risk, our findings might have differed if we had exclusively examined this subset. However, recent studies have indicated that the mode of antipsychotic discontinuation is less important than previously believed (McCutcheon et al., Reference McCutcheon, Taylor, Rubio, Nour, Pillinger, Murray and Jauhar2023; Moncrieff et al., Reference Moncrieff, Crellin, Stansfeld, Cooper, Marston, Freemantle and Priebe2023). Fourth, data collection always includes subjective elements. However, we have addressed this by implementing defined algorithms in cases of doubt, ensuring a transparent and rigorous data collection process. Finally, statistical guidelines normally recommend 10 or more outcome events per predictor variable in Cox regression models (Vittinghoff & McCulloch, Reference Vittinghoff and McCulloch2007). Our multivariate analyses, with eight predictor variables each, yielded 4 and 7.4 outcomes per variable, respectively, and accordingly they fall short of the ideal threshold. Despite this statistical concern, the univariate analyses showed mainly the same results as the multivariate analyses. Thus, there is little reason to anticipate different outcomes with a reduced number of predictor variables in the multivariate analyses.

A major strength of this study is the large and unselected sample. Due to the study design and the authorization to include all patients discharged with a diagnosis of schizophrenia from a psychiatric acute unit, irrespective of consent, the sample includes even the most severely ill patients. Accordingly, the sample is highly representative for patients with schizophrenia admitted to a psychiatric acute ward. A less ill group with higher level of insight might, paradoxically, be at higher risk of suicide. Hence, it is possible that in a sample of less severely ill patients than ours, the associations between the use of psychotropic drugs and the risk of suicidality might be different from those found in our study. The time-dependent analyses, reflecting a real-life setting where patients use psychotropic medications intermittently, contribute further to high generalizability. In contrast, antipsychotic drug trials in general, and in particular randomized controlled trials, have faced criticism for their limited generalizability due to issues like selection bias, short trial durations, and small sample sizes (Leucht, Heres, Hamann, & Kane, Reference Leucht, Heres, Hamann and Kane2008). Moreover, the study has a long follow-up period of up to 10 years. Data on suicides were gathered from the Norwegian Cause of Death Registry and there is only one psychiatric acute ward in the catchment area, both contributing to complete follow-up data. The study also accounts for potentially confounding factors such as alcohol or substance abuse, depression, and aggression. As such, the study will probably provide a more robust and less biased analysis of the association between suicidality and the use of different psychotropic drugs in a population with sufficiently severe symptoms to require acute hospitalization.